Research on Miners' Human Posture Detection Algorithm Based on MMPose

In UAV-based human detection, the extraction and selection of the feature vector are one of the critical tasks to ensure the optimal performance of the detection system. Although UAV cameras capture high-resolution images, human figures' relative size renders persons at very low resolution and contrast. Feature descriptors that can adequately discriminate between local symmetrical patterns in a low-contrast image may improve a human figures' detection in vegetative environments. Such a descriptor is proposed and presented in this paper. Initially, the acquired images are fed to a digital processor in a ground station where the human detection algorithm is performed. Part of the human detection algorithm is the GeHOG feature extraction, where a bank of Gabor filters is used to generate textured images from the original. The local energy for each cell of the Gabor images is calculated to identify the dominant orientations. The bins of conventional HOG are enhanced based on the dominant orientation index and the accumulated local energy in Gabor images. To measure the performance of the proposed features, Gabor-enhanced HOG (GeHOG) and other two recent improvements to HOG, Histogram of Edge Oriented Gradients (HEOG) and Improved HOG (ImHOG), are used for human detection on INRIA dataset and a custom dataset of farmers working in fields captured via unmanned aerial vehicle. The proposed feature descriptor significantly improved human detection and performed better than recent improvements in conventional HOG. Using GeHOG improved the precision of human detection to 98.23% in the INRIA dataset. The proposed feature can significantly improve human detection applied in surveillance systems, especially in vegetative environments.

Background: This study aims to model the kinematics of human joints using the Denavit-Hartenberg matrix method (hereinafter referred to as D-H matrix method) and combine biomechanical analysis for posture evaluation, thereby providing a more accurate and efficient detection solution. It ensures the implementation of complex calculations under low-power conditions and has broad application prospects in fields such as rehabilitation medicine, sports analysis, and virtual reality. Objective: The aim of this study is to design a sensor fusion-based embedded electronic system by integrating nine-axis sensors such as accelerometers, gyroscopes, and magnetometers. This system combines the D-H matrix method and forward kinematics for human posture detection and biomechanical analysis, to improve the system’s detection accuracy and response speed. Methods: Traditional forward kinematics and the D-H matrix method are used for kinematic modeling to enhance the accuracy and efficiency of posture calculation. Innovation: The D-H matrix method, a classical analysis technique in robotics typically used for kinematic analysis of robotic arms, is successfully applied in this study to human posture detection, breaking through traditional posture analysis methods. By utilizing the D-H matrix method to model the movement relationships between human joints, this study provides a more precise mathematical model for posture detection. By combining embedded electronic systems with biomechanical analysis to evaluate human posture, and introducing real-time monitoring of biomechanical loads from a biomechanical perspective, this study ensures that real-time human posture detection is not only efficient but also capable of performing complex calculations under low power conditions. Results: To further improve the accuracy of the sensors, this study analyzed the error characteristics of the inertial sensors and applied preprocessing algorithms to correct the errors in the signals from the magnetometer, accelerometer, and gyroscope. Combined with a high-pass and low-pass complementary filter fusion algorithm, the experiment showed that this algorithm successfully resolved the random drift and cumulative errors in the attitude angles calculated. The posture calculation system using the D-H matrix method outperforms the traditional forward kinematics method in terms of response time and root mean square error (hereinafter referred to as MSE). For instance, the response time for the right upper arm is reduced by 74.67% compared to traditional methods, while the MSE remains within a reasonable range.

This paper briefly describes the similarities and differences of the mainstream models of deep learning target detection box, analyzes the characteristics and advantages of Mask RCNN, a universal target detection box, and focuses on the application of Mask RCNN in human posture detection in multi- person human posture task. Through the analysis, it is considered that the advantage of Mask RCNN in multi-person human posture detection task is the accuracy, while the bottleneck is the detection speed. To solve this problem, an optimization of Mask RCNN model based on MobileNet was proposed to accelerate the inference calculation speed of Mask RCNN. At the same time, in order to further improve the detection accuracy of Mask RCNN, a method of using pixel segmentation results to assist the detection of human body key points is proposed. Experimental results show that compared with the original algorithm, it improves the reasoning speed and reduces the false detection rate caused by the environment.

Video surveillance involves petabytes of data storage requiring expensive hardware, which might also be time-inefficient. The aim of this article is, therefore, to develop an intelligent system capable of analyzing long sequences of videos captured from CCTV, helping to mitigate catastrophe and mitigate the violent threats faced by citizens every day, economically and efficiently. Existing models have achieved high accuracy on available datasets, the primary focus is to improve speed (time-efficient) of violence detection and use very little storage (economical) such that the system can be used in real-time. The paper presents an end-to-end hybrid solution for detecting violence in real-time video frames incorporating both human and weapon detection algorithms applied in a synchronized way. The focus of this article is to propose a generic HWVd (Human Weapon Violence detection) model to detect all kinds of public violence. HWVd is a three-tier ensemble model to detect violence in videos. The first tier is human detection, which uses a LightTrack framework. In the second tier, a Fast Region-based Convolutional Neural Network (F-RCNN) to detect any weapon in videos is used. The third tier uses a pre-trained VGG 19 (a pre-trained model of CNN) for spatial feature extraction and Long Short Term Memory (LSTM) to detect violent activity. Lastly, the output of this framework is sent to the Support Vector Machine to classify the activity as (i) violence not involving weapon, (ii) violence involving weapon and (iii) non-violent. The accuracy obtained using the proposed model is 98%.

Human posture detection is easily affected by the external environment, resulting in blurred results of limb feature extraction. In order to improve the accuracy and speed of human motion detection, this paper proposes a deep learning-based motion detection method in competitive sports training. The double parallel convolution network algorithm in the depth learning algorithm is used to process the collected action information, extract the body action features, and greatly reduce the operation scale; With the help of the theory of motion mechanics, the mechanical parameters in the motion process are calculated to eliminate outliers and reduce feature dimensions; With the help of motion inertial sensors and joint degrees of freedom, the limb motion detection results are obtained. The experimental results show that the average recognition rate of the method for different motion actions is 99.5%, and the average detection time is 148 ms, with good application performance.

This paper presents an algorithm for human presence detection in urban environments using an ultra-wide-band (UWB) impulse-based mono-static radar. A specular multi-path model (SMPM) is used to characterize human body scattered UWB waveforms. The SMPM parameters are used within a classical likelihood ratio detector framework to detect the presence of humans via gait, with the aid of a multi-target tracking technique (MTT). Experimental results on a simple human gait detection problem in an outdoor urban environment are presented to illustrate and validate the approach.

Aiming at the problems of low intelligence and energy waste in the current university library lighting system, this paper designs an intelligent lighting solution based on sub-area autonomy control. The improved YOLOv5 human target detection algorithm and the human spatial localization algorithm based on multilayer perceptron are used. The introduction of a coordinate attention mechanism and depth-separable convolution into the YOLOv5 backbone network, along with the optimization of the bounding box regression loss function, resulted in a mean average precision (mAP) of 91.4% for human body detection. The MLP model is used to map the 2D coordinates of the human body output from the human target detection algorithm to the 3D coordinates in the actual space, which realizes the fast and accurate localization of the distribution of multiple people in the museum. The related hardware circuits are designed to build a simple lighting system and simulated experiments. The experimental results show that the mean average accuracy of the proposed algorithm increased by 6.9% points, and the program can effectively complete the task of human body detection and localization, which provides certain reference significance for the research of intelligent lighting system in college libraries.

Human Posture detection using Deep learning techniques are widely useful in many areas like face recognition, video games, virtual reality, animations and many varieties of physical training. In this work, we are applying deep learning techniques like convolution neural networks (CNN) in the field of Yoga. It is a form of physical exercise which helps to improve both the physical and mental state of humans. There are different styles of yoga that combine physical poses, breathing techniques, and meditation or relaxation to promote mental and physical well-being. We have used a dataset which consists of five different yoga poses for training the model and evaluating the prediction accuracy of the model. The system recognizes a yoga pose from a given image using a CNN model and three CNN architectures such as VGG16, VGG19 and MobileNet. In comparison, among the four, VGG19 model achieved an accuracy of 99.49% which is higher than the other models.

This Human Skin Colour detection is analyzed to be a most suitable and strong cue for face detection and face recognition which finds various application in domain like surveillance, face biometrics, gesture recognition, interactive game application and many other. These applications require fast image processing in real time and demands enormously high performance with respect to time and processing speed. Earlier work done on sequential architecture does not provide required capability and can be achieved by parallel programming. As Image processing applications shows high degree of parallelism, they prove excellent source for multi-core platform. We can use GPU multithreaded parallel computing techniques to improve the speed of image processing. The paper proposes a data parallelism programming model for Human Skin Colour Detection algorithm. The objective is to increase the computational speed of the algorithm through data parallelism using CUDA framework. The framework is supported by OpenCV libraries and implemented by GPU (Graphics Processing Unit). The evaluation is done on basis of comparative analysis of serial and parallel programming computing. The pixel based Explicitly Defined Region skin classifier is used and RGB colour space is chosen due to its wider use in storing and processing of digital image data. To gain substantial acceleration in image computations, skin classifier code is off-loaded to GPU having compute capacity of 3.5 and rest of code is executed on CPU. We have tested module on static images, as well as on live camera captured images. SFA standard dataset is utilized to evaluate performance. Speedup of 23.45% was achieved using parallel programming.

Aim: To assess the relationship between serum ceramides and the main cardiovascular (CVD) risk factors in coal mining industry workers. Material and methods: The single-center study included 209 Kuzbass mining industry workers aged 39.0 (34.0; 45.0) years. The median length of service in underground mining was 14.0 (10.0; 16.0) years. The traditional cardiovascular risk factors were evaluated and the ceramide content Cer (d18:1) with a fatty acid residue with a different length of the hydrocarbon chain was analyzed. Results: The results revealed high rates of smoking (73.2 %), arterial hypertension (AH) (63.3 %), alcohol consumption (54.1 %), dyslipidemia (64.6 %) and obesity (24.9 %) in coal industry workers. Hypercholesterolemia (60.3 %) and elevated levels of lowdensity lipoproteins (LDL-С) (34.5 %) were the most common variant of dyslipidemia. Moreover, we detected low levels of high-density lipoproteins (HDL-С) (23.9 %) and hypertriglyceridemia (23 %) in coal industry workers with the same frequency. Respondents over 40 years of age showed worse clinical profile compared to younger participants due to a higher number of cases with hypertension, higher systolic (SBP) and diastolic (DBP) blood pressure, plasma cholesterol (TC). In addition, older respondents presented with hyperglycemia 2 times more often and with hypercholesterolemia 1.4 times more often compared to younger respondents. Taking into account the age of participants, comparative analysis of ceramides did not reveal any statistically significant differences. The length of service in the industry was characterized by a higher age of the respondents, a higher frequency of hypertension, as well as higher values of DBP and concentration of TC. There were no differences in the frequency of dyslipidemia and its variants. The content of Cer (d18:1/21:0) increased with length of service, and the content of Cer(d18:1/24:0) had an inverse relationship and decreased with increasing length of service in underground mining. Thus, the levels of Cer(d18:1/12:0) and Cer(d18:1/22:0) were associated with a high risk of obesity, hypertension, dyslipidemia, hypercholesterolemia, low HDL, high LDL, and hypertriglyceridemia. The maximum number of ceramides studied was associated with the risk of developing high LDL levels. Conclusions. Findings showed high frequency of traditional CVD risk factors such as AH, smoking, alcohol consumption, dyslipidemia and obesity in coal industry workers. The concentration of the studied ceramides did not depend on the age of workers. Longer length of service in the coal industry was associated with an increase in plasma concentrations of ceramides such as Cer(d18:1/14:0), Cer(d18:1/18:0), Cer(d18:1/20:0), Cer d18:1/21:0, and a decrease in Cer(d18:1/24:0). Cer (d18:1/24:0) was not associated with any of the parameters characterizing lipid metabolism disorders. Cer(d18:1/12:0) and Cer(d18:1/22:0) levels were associated with a high risk of obesity, hypertension, dyslipidemia, hypercholesterolemia, low HDL-С, high LDL-С and hypertriglyceridemia.

Deepfakes are synthetic media created by deep-generative methods to fake a person's audio-visual representation. Growing sophistication of deepfake technology poses significant challenges for both machine learning (ML) algorithms and humans. Here we used real and deepfake static face images (Study 1) and dynamic videos (Study 2) to (i) investigate sources of misclassification errors in machines, (ii) identify psychological mechanisms underlying detection performance in humans, and (iii) compare humans and machines in their classification decision accuracy and confidence. Study 1 found that machines achieved excellent performance in classifying real and deepfake images, with good accuracy in feature classification. Humans, in contrast, experienced challenges in distinguishing between real and deepfake static images. Their classification accuracy was at chance level, and this underperformance relative to machines was accompanied by a truth bias and low confidence for the detection of deepfake images. Using dynamic video stimuli, Study 2 found that performance of machines was near chance level, with poor feature classification. Further, machines showed greater lie bias and reduced decision confidence relative to humans who outperformed machines in the detection of video deepfakes. Finally, Study 2 revealed that higher analytical thinking, lower positive affect, and greater internet skills were associated with better video deepfake detection in humans. Combined, the findings across these two studies advance understanding of factors contributing to deepfake detection in both machines and humans; and can inform intervention toward tackling the growing threat from deepfakes by identifying areas of particular benefit from human-AI collaboration to optimize the detection of deepfakes.

Compared to 25 years ago, the climbing sport itself has changed dramatically. From a rock climbing modification to a separation in three independent disciplines, the requirements to athletes and trainers increased rapidly. To ensure continuous improvement of the sport itself, the usage of measurement and sensor technology is unavoidable. Especially in the field of the discipline speed climbing, which will be performed as a single discipline at the Olympic Games 2024 in Paris, the current state of the art of movement analysis only consists of video analysis and the benefit of the experience of trainers. Therefore, this paper presents a novel method, which supports trainers and athletes and enables analysis of motion sequences and techniques. Prerecorded video footage is combined with existing feature and human body keypoint detection algorithms and standardized boundary conditions. Therefore, several image processing steps are necessary to convert the recorded movement of different speed climbing athletes to significant parameters for detailed analysis. By studying climbing trials of professional athletes and the used techniques in different sections of the speed climbing wall, the aim among others is to get comparable results and detect mistakes. As a conclusion, the presented method enables powerful analysis of speed climbing training and competition and serves with the aid of a user-friendly designed interface as a support for trainers and athletes for the evaluation of motion sequences.

With the development of automation in the ports, more and more attention has been paid to the effective real-time monitoring of personnel intrusion into the unmanned area in the process of operation. In this paper, a fast human detection algorithm is proposed. Firstly, by improving HOG algorithm flow, the speed of feature extraction of HOG is greatly accelerated. Then, based on the HOG features, a 2-stage classifier based on Adaboost is proposed, which trains the front/back and side of the human sample library, so that the algorithm can adapt to the multi-pose human shape detection under the complex ports backgrounds. Finally, this paper presented a group of experiments about human detection on the container reach stacker of Shanghai International Port(Group) YIDONG Container Terminal Branch. The results show that the recognition accuracy of the combined algorithm of fast HOG and 2-stage classifier can reach 95% , and the detection time can be within 150ms. It realizes the calculation of 5-channel video human detection at the same time in a small embedded board, and meets the security requirements of unmanned areas in the ports.

Nation aging raises a problem of elderly health care and this problem will only grow stronger next decades. Preventive and remote care for seniors come to the fore as the solution to lower the number of potential patients in clinics and to reduce the cost of medication in case of accidents by providing first aid in time. Human fall is a source of danger in any age and, especially for older people. It can become a cause of injury itself and it can also be a result of dangerous event in human organism. That's why human fall detection is significant. This paper describes a first phase of research dedicated to developing a fall detection algorithm for a wrist-worn wearable device. Comparison and analysis of existing algorithms and solutions on market is made. Hardware setup for human fall detection is built and tested. Live tests on volunteers are performed to evaluate accuracy of chosen type of algorithm. Also algorithm is tested during daily activity to determine most fall-similar actions, and work out a way to distinguish those from true fall.

In order to make the elderly care service system intelligent, this paper proposes a human posture recognition algorithm based on big data analysis and artificial intelligence technology. The algorithm framework includes three parts: human target detection, human posture estimation and human action recognition. Firstly, this paper adopts the obscured human body detection algorithm based on the mixed attention mechanism to detect the human bounding box. Then, to solve the problem that the detected human bounding box may have inaccurate positioning leading to the human body position deviation, the space transformation network is used to affine the coordinates of the human bounding box to correct its position. Finally, the stacked hourglass network is used to predict the joints of the detected human body, so as to realize the recognition of human posture. In this paper, the algorithm is trained and tested on the open pedestrian test set. The experimental results show that the algorithm in this paper can effectively and correctly estimate the visible parts of human posture, and reduce the influence of obstacle occlusion on human posture estimation and human fall detection. The evaluation index of human posture recognition (Kps AP) reached 68.7%, indicating that the posture recognition algorithm proposed in this paper can realize the rapid recognition and estimation of human posture without shelter, and the algorithm can complete the fall recognition of the elderly in the elderly service system, and can effectively prevent the elderly from falling after the subsequent big data statistics and analysis of the elderly's posture. It has good practicability.